Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 48, 1, pp. 71-86, Warsaw 2010 EXPERIMENTAL MODEL OF FRACTURE OF FUNCTIONALLY GRADED MATERIALS Mieczysław Jaroniek Technical University of Lodz, Department of Materials and Structures Strength, Lodz, Poland e-mail: mieczyslaw.jaroniek@p.lodz.pl Fabrication of functionally gradedmaterials (FGM) can be obtained by layeredmixingof twomaterials ofdifferent thermo-mechanicalproperties with different volume ratios gradually changing from layer to layer such that the first layer has only a few particles of the second phase and the last has the maximum volume ratio of the first phase. Consider a simplemodel of the functionally gradedmaterials as amulti- layered beam bonded to planes having shear modulus Gi and Poisson’s ratio νI, respectively, subjected to bending. The behaviour of cracks de- pends on cracks configuration, size, orientation,material properties, and loading characteristic. The fracture mechanics problem will be appro- ached by making use of photoelastic visualisation of fracture events in the model structure. Key words: composites, fracture mechanics, photoelastic method, finite element method, functionally gradedmaterials 1. Introduction AN FGM material has functionally graded thermal and stress barriers. Be- tween the 100% ceramic layer and 100% metallic bond layer there exist a functionally graded layer 2 that contains some volume ratio of the bond (me- tallic) phase Vb(y), as a function of the distance y from thebond layer and the rest volume ratio of the ceramic phase Vc(y), also as a function of distance y from the bond layer, such that Vb(y)+Vc(y)= 1, where h is the thickness of the functionally graded layer. Fabrication of theFGMcan be obtained similarly by layeredmixing of two photoelasticmaterials of different thermo-mechanical propertieswith different volume ratios gradually changing from layer to layer such that the first layer 72 M. Jaroniek has only a few particles of the other phase and the last has the maximum volume ratio of the first phase. The development of the failure criterion for a particular application is also very important for predictions of the crack path and critical loads. Recently, there has been a successful attempt in formulation of problems of multiple cracks without any limitation. This attempt was concluded with a series of papers summarising the undertaken research for isotropic (Cook andErdogan, 1972), anisotropic (Gupta, 1973) and non-homogeneous class of problems, see Hilton and Sin (1971), Gupta (1973). Crack propagation in multi-layered composites of finite thickness is espe- cially challenging and still remains an openfield for investigation. Some results have been recently reported in Hilton and Sin (1971). The numerical calcu- lations were carried out using the finite element programs ANSYS 9 and 10 (User’s..., 2006). Two differentmethods were used: solid modelling and direct generation. 2. Material properties Material properties influence the stressdistributionandconcentration, damage process and load carrying capacity of elements. In the case of elastic-plastic materials, a region of plastic strains originates in most heavily loaded cross- sections. In order to visualise the state of strains and stresses, some tests have been performed on samplesmade of the ”araldite”-type optically active epoxy resin (Ep-53) modified with softening agents in such a way that an elastic material has been obtained. Properties of the components of the experimental model are given in Table 1. Table 1.Mechanical properties of components of the experimental model Layer Young’s Poisson’s Photoelastic Photoelastic modulus ratio constants in constants in Ei νi terms of stress terms of strain [MPa] [–] kσ [MPa/fr] fε [1/fr] 1 3450.0 0.35 1.68 6.572 ·10−4 2 1705.0 0.36 1.18 9.412 ·10−4 3 821.0 0.38 0.855 14.31 ·10−4 4 683.0 0.40 0.819 16.79 ·10−4 Experimental model of fracture... 73 3. Experimental results Dimensions of a typical model used in the experiment and artificially initiated small cracks in the tension zone are given in Fig.1a. Photoelastic models before test (under initial loading) in a circular pola- riscope and in a monochromatic sodium light are presented in Fig.1b, from which we get isochromatics in the layer in the initial phase. Isochromatics are the locus of points alongwhich the difference in the first and second principal stress (σ1−σ2) remains the same. Further, as the load was increased, the isochromatics ran parallel to the beam axis Figure 1c gives a picture of stress distribution even around abrupt discontinuities in the material (in the neighbourhood of the crack). Fig. 1. (a) Four-layer beamwith cracks. Photoelastic model under four point bending, isochromatic patterns (σ1−σ2) distribution. (b) Initial loading (P =20.0N); (c) P =50.0N – tension of layers 2, 3 and 4 The stress distribution was determined by making use of two methods: Shear Stress Difference Procedure (SDP – evaluation of the complete stress state by means of the isochromatics and angles of isoclines along the cuts) (Frocht, 1960). 74 M. Jaroniek The photoelastic model of a four layer beamwith cracks under four point bending, the isochromatic patterns (σ1−σ2) distribution corresponding to ini- tiation of the vertical cracks and tension of overcoat (layer 4), and compression of the substrate (layer 1) are presented in Fig.2. Fig. 2. Experimentally obtained isochromatic patterns (σ1−σ2) according to the initiation of crack propagation Thestressdistributioncorrespondingto initiation of thevertical crackswas evaluated frommeasurement results of the isochromatic and isoclinic patterns according to Frocht (1960). By means of the angles of isoclinics of isochromatics along the cuts y et y + ∆y and by employing the equations of equilibrium, the complete stress state was determined. Thestressdistribution σx in cross sections A-Aand B-B determinedusing the shear stress difference procedure, by means of the angles of isoclinics, is shown in Fig.3. Method of characteristics (stress distribution was determined using iso- chromatics only and equations of equilibrium (Szczepiński, 1961). In a general case (Sanford and Dally, 1979), the Cartesian components of stress: σx, σy and τxy in the neighbourhood of the crack tip are σx= 1√ 2πr Bigl[KI cos Θ 2 ( 1−sinΘ 2 sin 3Θ 2 ) −KII sin Θ 2 ( 2+cos Θ 2 cos 3Θ 2 )] +σox σy = 1√ 2πr [ KI cos Θ 2 ( 1+sin Θ 2 sin 3Θ 2 ) +KII sin Θ 2 cos Θ 2 cos 3Θ 2 ] (3.1) τxy = 1√ 2πr [ KI sin Θ 2 cos Θ 2 cos 3Θ 2 +KII cos Θ 2 ( 1− sin Θ 2 sin 2Θ 2 )] Experimental model of fracture... 75 Fig. 3. Distribution of stresses in cross sections A-A and B-B (0.5mm) with respect to the crack according to experiment and SDP evaluation From which (σ1−σ2)2 = 1 2πr [(KI sinΘ+2KII cosΘ) 2+(KII sinΘ) 2]+ (3.2) −2 σox√ 2πr sin Θ 2 [KI sinΘ(1+2cosΘ)+KII(1+2cos 2Θ+cosΘ)]+σ2ox where KI and KII are the stress-intensity factors, r and Θ are coordinates in thepolar coordinate system.By inserting thevalues kσmi = σ1−σ2 into (3.2), we obtain isochromatic curves in polar coordinates (r,Θ). For each isochro- matic loop, the position of maximum angle Θm corresponds to themaximum radius rm. Thisprinciple can alsobeused in themixedmodeanalysis (Sanford and Dally, 1979) by employing information from two loops in the near field of the crack, if the far field stress component σox(Θ)= const. Differentiating Eq. (3.2) with respect to Θ, setting Θ = Θm and r = rm, and substituting that ∂τm/∂Θm =0, gives g(KI,KII,σox)= 1 2πr (K2I sin2Θ+4KIKII cos2Θ−3K2II sin2Θ)+ −2 σox√ 2πrr sin Θ 2 { [KI(cosΘ+2cos2Θ)−KII(2sin2Θ+sinΘ)]+ + 1 2 cos Θ 2 [KI(sinΘ+sin2Θ)+KII(2+cos2Θ+cosΘ)] } f(KI,KII,σox)= (σ1−σ2)2− (kσm)2 =0 and g(KI,KII,σox)= ∂(σ1−σ2)2 ∂Θm =0 (3.3) 76 M. Jaroniek Substitution of the radii rm and the angles Θm from these two loops into a pair of equations of the formgiven inEq. (3.3) gives two independent relations for the parameters KI, KII and σox. The third equation is obtained by using Eqs. (3.2). The three equations obtained that way have the form gi(KI,KII,σox)= 0 gj(KI,KII,σox)= 0 (3.4) fk(KI,KII,σox)= 0 In order to determine KI, KII and σox, it is sufficient to select two arbitrary points ri, Θi and apply the Newton-Raphsonmethod to the solution of three simultaneous non-linear equations (3.4). The values KC according to mixed mode of the fracture were obtained from KC = √ K2I +K 2 II (3.5) Below are exemplary numerical results obtained from (3.4) m =12.5 r1 =0.6mm Θ1 =1.484 r2 =10.45mm Θ2 =1.416 K (4) I =0.14MPa √ m K (4) II =1.05MPa √ m σox =0.039MPa K (4) C =1.05MPa √ m By inserting ri, Θi in the three selected arbitrary points into (3.2), we obtain three non-linear equations (i =1,2,3) fi(KI,KII,σox)= 0 (3.6) and applying the Newton-Raphson method to the solution, we have KI, KII and σox. Exemplary numerical results (shown in Fig.3) obtained from (3.6) are as follows m1 =12.5 r1 =0.72mm Θ1 =1.484 m2 =8.0 r2 =1.15mm Θ2 =1.37 m3 =5.5 r3 =1.85mm Θ3 =1.315 and K (4) I =0.702MPa √ m K (4) II =1.043MPa √ m σox =0.152MPa K (4) C =1.257MPa √ m Experimental model of fracture... 77 4. Stress analysis from complex potentials In the present paper, the elastic and plastic deformation has been approxima- tely analysed usingMuskhelishvili’s complex potentials method σxx+σyy =4Re[ψ ′(z)] σyy −σxx+2iτxy =2[zψ′′(z)+χ′′(z)] 2G(u+iυ)= κψ(z)−zψ′(z)−χ′(z) 2G(u′+iυ′)= κψ′(z)−zψ′(z)−zψ′′(z)−χ′′(z) where κ = 3−4ν for the plain strain and κ = (3−ν)/(1+ν) for the plain stress σxx+σyy =4Re[ψ ′(z)] σyy −σxx =Re2[zψ′′(z)+χ′′(z)] τxy =Im[zψ ′′(z)+χ′′(z)] The elastostatic stress field is required to satisfy the well-known equilibrium equations (Cherepanov, 1979) using two analytical functions ψ(z) and χ(z) σxx =Re[2ψ ′(z)−zψ′′(z)−χ′′(z)] σyy =Re[2ψ′(z)+zψ′′(z)+χ′′(z)] τxy =Im[zψ ′′(z)+χ′′(z)] χ′′(z)=−zψ′′(z) (4.1) 2G(u+iυ)= κψ(z)−zψ′(z)−χ′(z) Using these two analytical functions ψ(z) and χ(z) ψ′(z)= ψ′1(z)+ψ ′ 2(z) χ ′′(z)= z[ψ′′1(z)−ψ′′2(z)] The stresses are assumed as follows σx =Re[2ψ ′−2xψ′′1]−yIm2ψ′′2 σy =Re[2ψ′+2xψ′′1]+yIm2ψ′′2 (4.2) τxy = xIm2ψ ′′ 1 −yRe2ψ′′2 The stress-intensity factors are related by KI = lim x→a σy(x,0) √ 2π(x−a) KII = lim x→a τxy(x,0) √ 2π(x−a) (4.3) By inserting (4.2) into (4.3), we obtain K (j) I = limx→a {Re[2ψ′(j)(z)]+x2ψ′′(j)(z)} √ 2π(x−a) K (j) II = limx→a {xIm[2ψ(j)(z)]} √ 2π(x−a) 78 M. Jaroniek The complex stress potentials are assumed as follows 2ψj(z)= N ∑ n=1 ( Cn ∫ z(2−n) √ z −a z +a dz ) (4.4) 2ψ ′(j)(z)= N ∑ n=1 √ z −a z +a Cjnz (2−n) In this case, theCartesian components of the stresses σx, σy and τxy are given as σjx = N ∑ n=1 { Re [ (C j n+iC j n)f1(z) ]} σjy = N ∑ n=1 { Re [ (C j n+iC j n)f4(z) ]} (4.5) τjxy = N ∑ n=1 { xIm [ (C j n+iC j n)f5(z) ]} where f1(z)= 1 zn−1 (1 z √ z −a z +a −x az− (n−2)(z2−a2) (z+a) √ z2−a2 ) f2(z)= z3−n√ z2−a2 f3(z)= z 2−n(3−n)(z2−a2)+z2 (z2−a2)3/2 f4(z)= 1 zn−1 (1 z √ z −a z +a +x az− (n−2)(z2−a2) (z+a) √ z2−a2 ) f5(z)= 1 zn−1 az(n−2)(z2−a2) (z +a) √ z2−a2 and Cin =ReC i n+iImC i n C i n =ReC i n Din =ReD i n+iImD i n C i n =ImC i n (4.6) The boundary conditions can be expressed as σ(∞)y −σ(∞)x +2iτ(∞)xy =2iy ·2ψ(z) (4.7) σjy = Ej (x0−xj) ρ = σ∞y σ ∞ x =0 Experimental model of fracture... 79 εjy = ε j+1 y ε j y = ∂υj ∂y εjx = ε j+1 x u′ = ∂u ∂x uj = uj+1 υj = υj+1 (4.8) and H ∫ 0 σjy dx = N ∑ n=1 { Re H ∫ 0 [ Cjn zn−1 (1 z √ z−a z+a +x az − (n−2)(z2 −a2) (z +a) √ z2−a2 )] dx } =0 H ∫ 0 bσjyx dx+M = = N ∑ n=1 { H ∫ 0 bRe [ Cjn zn−1 (1 z √ z−a z+a +x az − (n−2)(z2−a2) (z +a) √ z2−a2 )] dx } +M =0 (4.9) σjy = N ∑ n=1 { Re [ Cjn zn−1 (1 z √ z−a z+a +x az − (n−2)(z2−a2) (z +a) √ z2−a2 )]} σjy = Ej (x0−xi)p ρ =σ∞y For each isochromatic loop σ1−σ2 = kσm and σyy −σxx = kσmcos2α (4.10) in other hand σ1+σ2 = σxx+σyy and σxx+σyy = N ∑ n=1 2Re [ Cnz (2−n) √ z −a z +a + Dnz (3−n) √ z2−a2 ] (4.11) This principle can be used in the analysis of stresses. By inserting values of mi, αi in arbitraily selected points P(ri,Θi), we obtain equations N ∑ n=1 { Re [ (Cn+iCn)(f4(z)−f1(z)) ] +2yIm [ (Dn+iDn)f3(z) ]} = kσmicos2αi (4.12) fromwhich σyy −σxx =Re2[zψ′′(z)+χ′′(z)] τxy =Im[zψ ′′(z)+χ′′(z)] τ2m =Re[zψ ′′(z)+χ′′(z)]2+Im[zψ′′(z)+χ′′(z)]2 80 M. Jaroniek and τ2m = 〈 N ∑ n=1 {xRe[Cnf5(z)]+yIm[Dnf3(z)]} 〉2 + (4.13) + 〈 N ∑ n=1 {xIm[Cnf5(z)]−yRe[Dnf3(z)]} 〉2 For each isochromatic loop εyy −εxx = (4.14) = 1 2G′ N ∑ n=1 { Re[(Cn+iCn)(f4(z)−f1(z))]+2yIm[(Dn+iDn)f3(z)] } in other hand εyy −εxx = fε(m)mcos2α (4.15) and (1+νj) Ej N ∑ n=1 { Re[(C j n+iC j n)(f4(z)−f1(z))]+2yIm[(D j n+iD j n)f3(z)]}= (4.16) = fεm(xt)cos2α By inserting the principal stress difference mi =(σ1−σ2)i, the values of αi in narbitrarily selectedpoints Pi(ri,Θi) into (4.16),weobtain n-linear equations (i =1,2, . . . ,n) fromwhichwe obtain values of C1,C2, . . . ,Cn,D1,D2, . . . ,Dn and from (4.7) and (4.8) Cartesian components of the stress σx, σy and τxy. By inserting Cn, Dn into (4.13), we obtain isochromatic curves in the polar coordinates (r,Θ). The distribution of isochromatic patterns (σ1 −σ2) obtained experimen- tally according to crack propagation is shown in Fig.4a and the isochromatics calculated from (4.13) according to Cn and Dn in the analysis of complex potentials are presented in Fig.4b. The stress-intensity factors from (9) are related by KI = lim x→a N ∑ n=1 { Re [ Cn zn−1 (1 z √ z−a z+a +x az − (n−2)(z2−a2) (z +a) √ z2−a2 ) + Dnz 3−n √ z2−a2 ] · · √ 2π(x−a) } (4.17) KII = lim x→a N ∑ n=1 { xIm [ Cn zn−1 az(n−2)(z2−a2) (z +a) √ z2−a2 ] √ 2π(x−a) } Experimental model of fracture... 81 Fig. 4. Experimentally found isochromatic patterns (σ1−σ2) according to crack propagation used in analysis of complex potentials where in the polar coordinates (r,Θ) zl √ z −a z +a = rl √ r1 r2 [ cos (Θ1+Θ2 2 +Θl ) +isin (Θ1+Θ2 2 +Θl )] (4.18) zl√ z2−a2 = rl √ r1r2 [ cos ( Θl− Θ1+Θ2 2 ) +isin ( Θl− Θ1+Θ2 2 )] 5. Numerical determination of stress distribution The distribution of stresses and displacements has been calculated using the finite element method (FEM) (User’s..., 2006; Zienkiewicz, 1971). Finite ele- ment calculations were performed in order to verify the experimentally obse- 82 M. Jaroniek rved isochromatic distribution during cracks propagation. The geometry and materials of models were chosen to correspond with specimens used at pre- sent experiments. The numerical calculations were carried out using the finite element programANSYS9 and by applying the substructure technique. A finite element mesh of the model (used for numerical simulation) are presented in Fig.5. Fig. 5. Finite elementmesh of the model (for numerical simulation) The numerical (from FEM) distribution of isochromatic fringes (σ1−σ2) is shown in Fig.6 and the distribution of stress σy obtained numerically is given in Fig.7. Fig. 6. Numerically obtained distribution of isochromatic patterns (σ1−σ2) according to crack propagation For comparison, the distribution of isochromatic fringes (σ1−σ2) obtained experimentally according to the initiation of crack propagation is shown in Fig.8. Experimental model of fracture... 83 Fig. 7. Numerical determination of stress distribution (Ansys9). Distribution of stresses σy along the crack Fig. 8. Experimentall obtained distribution of isochromatic patterns (σ1−σ2) according to crack propagation Table 2.Experimental and numerical results. Critical values K (4) IC and K (4) IIC according to crack propagation Crack Critical Experimental results Numerical length force results a Pcr K (4) I K (4) II K (4) C σ0x K (4) Cn [mm] [N] [MPa √ m] [MPa] [MPa √ m] 6.0 265.0 1.177 0.8793 1.419 2.58 1.45 9.0 205.0 0.702 1.043 1.257 0.152 1.28 9.8 185.0 0.14 1.05 1.05 0.039 1.16 84 M. Jaroniek 6. Conclusions Formulation of the model of an FGM can be obtained similarly as the func- tionally graded materials themselves by layered mixing of two photoelastic materials of different thermo-mechanical properties with different volume ra- tios gradually changing from one layer to another such that the first layer has only a few particles of the second one and the second has the maximum volume ratio of first layer. Photoelasticity has been shown to be promising in stress analysis of beams with various numbers and orientation of cracks. In the present paper, the stresses anddisplacements have been approxima- tely analysed using the Muskhelishvili complex potential method. Using two analytical functions, we obtained equations for the isochromatic loop. It is possible to create a model using various photoelastic materials to model a multi-layered structure. Finite element calculations (FEM) were performed in order to verify the experimentally observed branching phenomenon and the isochromatic distri- bution observed during crack propagation. The agreement between the di- stributions of isochromatic fringe patterns found numerically by FEM and determined photoelastically was found to be within 5-8 percent. Having used the stress-intensity-factor criterion, the critical value of stress and deformation fields characterises the fracture toughness. References 1. Cherepanov G.P., 1979, Mechanics of Brittle Fracture, McGraw-Hill, New York 2. Cook T.S., Erdogan F., 1972, Stresses in bonded materials with a crack perpendicular to the interface, International Journal of Engineering Science, 10, 677-697 3. Frocht M.M., 1960,Photoelasticity, JohnWiley, NewYork 4. Gupta A.G., 1973, Layered composite with a broken laminate, International Journal of Solids and Structures, 36, 1845-1864 5. Hilton P.D., Sin G.C., 1971, A laminate composite with a crack normal to interfaces, International Journal of Solids and Structures, 7, 913 6. Neimitz A., 1998,Mechanics of Fracture, PWNWarsaw [in Polish] Experimental model of fracture... 85 7. Rychlewska J., Woźniak C., 2006, Boundary layer phenomena in elasto- dynamics of functionally graded laminates, Archives of Mechanics, 58, 4/5, 431-444 8. Sanford R.J., Dally J., 1979, A general method for determining mixed- mode stress intensity factors from isochromatic fringe patterns,Engineering of Fracture Mechanics, 2, 621-633 9. Stupnicki J., 1965, Trends of experimentalmechanics, Journal of Theoretical and Applied Mechanics, 34, 2, 207-233 10. SzczepińskiW., 1961,A photoelasticmethod for determining stresses byme- ans isochromes only,Archiwes of Applied Mechanics, 13, 5 11. Szymczyk J., Woźniak C., 2006, Continuum modelling of laminates with a slowly gradedmicrostructure,Archives of Mechanics, 58, 4/5, 445-458 12. User’s Guide ANSYS, 2006, 9, 10, Ansys, Inc., Huston, USA 13. Woźniak C., 1993, Nonlinear macro-elastodynamics of microperiodic compo- sites,Bull. Ac. Pol. Sci.: Tech. Sci., 41, 315-321 14. Woźniak C., 1995,Microdynamics: continuum.Modelling the simple compo- site materials, Journal of Theoretical and Applied Mechanics, 33, 267-289 15. Zienkiewicz O.C., 1971,The Finite ElementMethod in Engineering Science, McGraw-Hill, London, NewYork Eksperymentalny model pękania materiału funkcjonalnie zmiennego Streszczenie Sposób wykonaniamateriałów funkcyjnie zmiennych – FGM, polega na nakłada- niu (lub napylaniu) kolejnych (możliwie cienkich) warstw. Warstwy te składają się zazwyczaj z dwóch składników o różnych własnościach (mechanicznych, termicznych itp.) np. ciekły metal (plazma) i ceramika. Technologia wykonania jest następująca: najpierw nakłada się warstwę czystego ciekłego metalu, następnie warstwę ciekłego metalu z niewielką domieszką ceramiki, następne warstwy to ciekły metal ze zwięk- szającą się ilością ceramiki, aż do ostatniej warstwy składającej się z czystej ceramiki. W ten sposób (po zastygnięciu) otrzymuje się powłoki izolacyjne stosowane do komór spalania silników, warstwy termoizolacyjne promów i kapsuł kosmicznych itp. W pracy podano metodę doświadczalno-obliczeniową stanu naprężeń i odkształ- ceńdla kompozytówwarstwowychzbudowanychzmateriałówgradientowych.Podano także dla takich kompozytów ze szczelinami kryterium zniszczenia. Opracowaname- toda jest metoda hybrydową – opierając się na wynikach badań doświadczalnych, opisano stan naprężenia i odkształcenia, stosując metodę potencjałów zespolonych. 86 M. Jaroniek Na podstawie wynikówbadań doświadczalnych opisano stan naprężeń i odkształ- ceń dlamateriałów liniowo-sprężystych i sprężysto-plastycznychw formie funkcji po- tencjałów zespolonych. Zastosowaniemetody potencjałów zespolonychw postaci sze- regów do analizy pól naprężeń umożliwia analizę wyników badań doświadczalnych modeli kompozytów, z uwagi na budowę równań, które w kompleksowy sposób opisu- ją stan naprężenia w formie sumy i różnicy naprężeń normalnych. Przeprowadzoneobliczenia i badaniamająna celuweryfikacjęmodelu, którymoż- na zastosować do obliczeń kompozytówwarstwowych. Manuscript received March 13, 2009; accepted for print April 20, 2009